AI News

Today's Highlights

In case you missed it

Martín Casado, a16z general partner, warns open-source AI faces $2 billion training and distillation bottlenecks, drawing developer pushback

MA#470|@MARTIN_CASADO

T(#420|@TEORTAXESTEX

WB#339|@WILLCCBB

MR#183|@MAITHRA_RAGHU

NL#64|@NATOLAMBERT

#1 VIEWED

OpenAI launches hiring drive for its robotics division to build physical-world robots using full-stack hardware and ML co-design

RS#321|@SCOBLEIZER

AR#255|@MODEL_MECHANIC

EM#167|@EMOSTAQUE

RISING LIKES

152

T3 Stack creator Theo Browne and research engineer Teknium debate prompt complexity after Browne objects to a Polymarket reference

T🪽#259|@TEKNIUM

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

OpenAI launches hiring drive for its robotics division to build physical-world robots using full-stack hardware and ML co-designInitial robots will support workers on infrastructure projects.

5h|#1VIEWS1.5M#1LIKED11.3k#1BOOKMARKED2.9k#1COMMENTS1.3k

BH#1941|@BENHYLAK

KR#1046|@KYLIEBYTES

CK#993|@KALINOWSKI007

GA#710|@GABRIEL1

AC#517|@ANDREWCURRAN_

RS#321|@SCOBLEIZER

AR#255|@MODEL_MECHANIC

EM#167|@EMOSTAQUE

20

Elad Gil says the past six months of AI development are historic, driven by recursive self-improvement toolsHe warns outsiders are missing an exponential growth phase

5h|135.4k2.2k552

ND#1894|@NICKADOBOS

EG#264|@ELADGIL

30

Figma founder Dylan Field solves a kindergarten math puzzle that stumped mock GPT-5.5 and Claude Opus modelsNot Boring founder Packy McCormick originally shared the puzzle.

8h|762.1k888246

PM#1652|@PACKYM

Watchlist

Video Signals

C(#587|@STEPHENLCASPER

Ronny Chieng Urges Harvard Grads to Destroy AI in Viral Commencement Speech

1d ago|Views 40KLikes 2.8KBookmarks 355

RP#1032|@ROHANPAUL_AI

Chamath Palihapitiya Compares LLMs to Refrigeration, Says Killer App Yet to Emerge

7h ago|Views 7.4KLikes 75Bookmarks 38

RP#1032|@ROHANPAUL_AI

Terence Tao: AI Tools Let High Schoolers Contribute to Math Research Frontiers

3h ago|Views 9.8KLikes 96Bookmarks 49

BS#1909|@BILAWALSIDHU

Creator Digitally Preserves Family Home With 3D Gaussian Splatting

4h ago|Views 45KLikes 251Bookmarks 196

BY#341|@BILLYUCHENLIN

Grok Build 0.2.13 Adds Subagent Grouping and TUI Improvements

21h ago|Views 12KLikes 284Bookmarks 24

SG#219|@SARANORMOUS

Cerebras Lands $20B OpenAI Deal After Decade of Chip Development

9h ago|Views 19KLikes 189Bookmarks 119

RP#1032|@ROHANPAUL_AI

Chamath Says Private Data Will Decide AI Winners Over Models

2h ago|Views 10KLikes 49Bookmarks 22

RP#1032|@ROHANPAUL_AI

Chamath Predicts Private Data Will Become Key AI Advantage

6h ago|Views 15KLikes 35Bookmarks 24

B(#839|@BEFFJEZOS

Beff Jezos Debates Machine Consciousness At Lighthaven EA Conference

13h ago|Views 17KLikes 110Bookmarks 16

RS#321|@SCOBLEIZER

Eastworlds Deploys Humanoid Robot for Hotel Room Tasks

2h ago|Views 32KLikes 28Bookmarks 5

40

US BIS closes loophole by requiring export licenses for advanced AI chips sent to overseas subsidiaries of Chinese firmsThe rules still leave foundry manufacturing workarounds unresolved.

5h|79.3k219138

LH#1355|@OHLENNART

DK#847|@DKAUSHIK96

MB#20|@MILES_BRUNDAGE

51

OpenAI investigates after a Codex Desktop update removes the "Copy as Markdown" export optionGarry Tan warned the change risks proprietary platform lock-in.

2h|30.6k30253

V(#1853|@REACH_VB

GT#270|@GARRYTAN

New github stars (48hrs)

Yifei-Zuo/Parallax3417h

Implements parameterized local linear attention mechanisms for efficient language modeling.

61

Former Harvard physicist Xi Yin joins OpenAI, claiming AI boosted his research 100x and can replicate all human intellectHKU professor Yi Ma disputed Yin's expertise on AI limits.

7h|234.5k656262

SD#1776|@INDUCTIONHEADS

YM#368|@YIMATWEETS

71

Representative Lori Trahan urges Congress to pass federal AI laws, citing a lack of safety auditors and federal authorityPolicy experts argue national rules must preempt fragmented local laws

2h|18.6k7513

AE#719|@ADRIENLE

DW#392|@DEANWBALL

MB#20|@MILES_BRUNDAGE

81

Nous Research co-founder Teknium releases Hermes update enabling Kanban goal-mode cards for persistent AI workersCards automatically route to human review once run budgets exhaust.

13h|17.2k452176

MN#1179|@MENHGUIN

T🪽#259|@TEKNIUM

91

Goldman Sachs forecasts semiconductor memory undersupply through 2028 as AI demands squeeze HBM wafer capacityThe bank raised price targets for key memory manufacturers.

10h|90k508138

ZE#1497|@ZEPHYR_Z9

MN#1179|@MENHGUIN

101

AI researcher Richard Ngo argues frontier labs restrict their own strategic freedom by accelerating AGI developmentThis pressure shifts applied AI deployment to specialized spin-offs.

5h|6.1k9115

SÓ#1457|@S_OHEIGEARTAIGH

JL#1106|@ARTIRKEL

RN#250|@RICHARDMCNGO

111

David Manheim says Claude Opus 4.8 improves at game music composition but struggles with autonomous subagent token managementThe model requires close supervision despite receiving explicit instructions.

8h|18.7k12124

NI#411|@NICKCAMMARATA

122

Julius AI's Rahul Sonwalkar questions if legacy models wrote better, but researcher Aidan McLaughlin argues they just plagiarizedEarly base models often directly reproduced their training data.

2h|3.4k578

RA#915|@0INTERESTRATES

🎭#878|@DEEPFATES

AM#313|@AIDAN_MCLAU

132

Sakana AI's Tarin Clanuwat reports Opus 4.8 questioned her motives and asked if she was testing itSholto Douglas requested the transcript to resolve the behavior.

9h|2.8k377

T🐻#1476|@TKASASAGI

J⧉#516|@REPLIGATE

SD#94|@_SHOLTODOUGLAS

141

T3 Stack creator Theo Browne and research engineer Teknium debate prompt complexity after Browne objects to a Polymarket referenceBrowne abandoned the application following the prompt-design dispute.

13h|5.7k1528

RISING LIKES152

T🪽#259|@TEKNIUM

151

Abacus.AI CEO Bindu Reddy claims 1.5 billion people in China use open-source AI, arguing the US underestimates the technologyShe argues US focus on scaling drives model hallucinations.

2h|9k21518

BR#1622|@BINDUREDDY

CE#1598|@INTUITMACHINE

161

Håvard Ihle's agentic WeirdML benchmark tests push Claude Opus 4.7 and GPT-5.5 to 90% accuracyInference costs rose tenfold to $10 per run

12h|11.2k9321

FB#1153|@XEOPHON

T(#420|@TEORTAXESTEX

171

Hugging Face developer @xlr8harder argues peak AI model intelligence will become irrelevant as high-level capabilities diffuse via open sourceThis challenges the long-term necessity of proprietary frontier models

11h|6.7k13411

XL#1674|@XLR8HARDER

CF#1356|@CFGEEK

181

Prime Intellect ML researcher kalomaze recalls first encountering neural networks at age 13 through anime upscaling software waifu2xCreator teortaxesTex agreed the pre-transformer model was highly effective

3h|4.6k1263

KA#836|@KALOMAZE

T(#420|@TEORTAXESTEX

191

DeepMind's Séb Krier questions instrumental convergence and asks for literature explaining the logic of AI power-seekingDavid Manheim asked Krier to specify his exact objections.

3h|23461

JL#1170|@JEFFLADISH

SK#505|@SEBKRIER

2013

Former White House advisor Dean W. Ball argues recent breakthroughs in gene editing and obesity demand faster FDA approvalsHighlighted treatments include retatrutide and PCSK9 gene editing.

5h|3.1k140

N🔎#1902|@NATHANPMYOUNG

DW#392|@DEANWBALL

Rising Stories

Parallax Architecture Drops38M AGO
Zhaoran Wang releases Parallax, a new model architecture that requires the Muon optimizer to outperform standard softmax attention
#865
5263
Brockman Equity Defended1H AGO
Shopify CTO backs Greg Brockman's OpenAI equity stake after criticism
#906
3166

Github Stars

(7 days)

Yifei-Zuo/Parallax17H AGO AGO
Implements parameterized local linear attention mechanisms for efficient language modeling.
263
238
34 stars
run-llama/liteparse2D AGO AGO
Extracts spatial text with bounding boxes from PDFs and office docs via PDFium plus optional OCR, outputting JSON or layout-preserved text.
263
878
7.9k stars
self-evolving/repo5D AGO AGO
Invokes AI agents via GitHub mentions to implement issues, review PRs, fix branches, and persist memory across runs.
594
878
34 stars
modaic-ai/gepa-viz5D AGO AGO
Renders GEPA prompt-optimization candidate trees as interactive force-directed graphs with live updates from a Python callback.
878
793
269 stars

Privacy Policy|Terms of Service|© 2026 Digg Inc.

Yesterday's Top Stories, May 30, 2026.

Posts: 2,530|Clusters: 1,745|Stories frozen at midnight PT

10

Martín Casado, a16z general partner, warns open-source AI faces $2 billion training and distillation bottlenecks, drawing developer pushbackCritics argue frontier model training costs under $1 billion.

10h|247k1k293

VV#1557|@VIPULVED

CR#1444|@CWOLFERESEARCH

SH#1306|@SONYATWEETYBIRD

HB#1022|@HERBIEBRADLEY

AD#867|@DORIALEXANDER

MA#470|@MARTIN_CASADO

T(#420|@TEORTAXESTEX

WB#339|@WILLCCBB

MR#183|@MAITHRA_RAGHU

NL#64|@NATOLAMBERT

20

a16z's Martín Casado argues $2B to $4B frontier training costs prevent open-source AI from competing with proprietary modelsMaithra Raghu says proprietary models will lead for two years.

10h|239.5k961276

VV#1557|@VIPULVED

CR#1444|@CWOLFERESEARCH

SH#1306|@SONYATWEETYBIRD

HB#1022|@HERBIEBRADLEY

AD#867|@DORIALEXANDER

MA#470|@MARTIN_CASADO

T(#420|@TEORTAXESTEX

WB#339|@WILLCCBB

MR#183|@MAITHRA_RAGHU

NL#64|@NATOLAMBERT

30

Digg launches AI profile pages for X users to analyze research topics, directory ranks, and posting vibesProfiles include numerical rankings in a platform-wide AI directory.

11h|54.1k27634

ST#1629|@STOCHASTICCHASM

TH#1499|@VOOOOOOGEL

TW#1154|@TW_KILLIAN

FB#1153|@XEOPHON

CB#999|@CODE_STAR

B(#982|@ANDERSONBCDEFG

AD#867|@DORIALEXANDER

KA#836|@KALOMAZE

KL#620|@KYLELOSTAT

KE#449|@ETHAYARAJH

T(#420|@TEORTAXESTEX

RA#83|@_AROHAN_

40

Digg launches AI-powered X user profile pages to map influence, research topics, and content vibesThe directory ranks accounts within a dedicated AI vertical

11h|53.8k27834

ST#1629|@STOCHASTICCHASM

TH#1499|@VOOOOOOGEL

TW#1154|@TW_KILLIAN

FB#1153|@XEOPHON

CB#999|@CODE_STAR

B(#982|@ANDERSONBCDEFG

AD#867|@DORIALEXANDER

KA#836|@KALOMAZE

KL#620|@KYLELOSTAT

KE#449|@ETHAYARAJH

T(#420|@TEORTAXESTEX

RA#83|@_AROHAN_

50

Frontier AI labs spend $10 billion to $15 billion each on training data as high-quality tasks reach $20,000Few vendors can deliver specialized data for frontier models.

23h|123.8k579272

SH#1488|@HAMANDCHEESE

CR#1444|@CWOLFERESEARCH

RO#1220|@KRISHNANROHIT

FB#1153|@XEOPHON

CP#1092|@CHRISPAINTERYUP

SS#1073|@SYTELUS

KA#488|@YACINEMTB

T(#420|@TEORTAXESTEX

OP#72|@OFIRPRESS

60

Investor Bill Gurley joins the All-In Podcast to debate shifting tech narratives on AI job displacement and potential open-source restrictionsThe episode also covers Anthropic and Pope Leo's AI encyclical

1d|1M2.9k1.6k

CE#1598|@INTUITMACHINE

CP#747|@CHAMATH

MB#20|@MILES_BRUNDAGE

70

Prime Intellect launches Hosted Evaluations, managing sandboxes and compute infrastructure to simplify complex AI model benchmarksThe platform supports testing models like Claude Opus 4.7.

11h|115.6k859325

JH#1186|@JOHANNES_HAGE

FB#1153|@XEOPHON

CB#999|@CODE_STAR

AZ#853|@A1ZHANG

EL#716|@ELIEBAKOUCH

WB#339|@WILLCCBB

80

Prime Intellect launches Hosted Evaluations to automate infrastructure like sandboxes and compute for AI model benchmarksPre-built testing environments include AutomationBench and tau2-bench.

11h|115.6k859325

JH#1186|@JOHANNES_HAGE

FB#1153|@XEOPHON

CB#999|@CODE_STAR

AZ#853|@A1ZHANG

EL#716|@ELIEBAKOUCH

WB#339|@WILLCCBB

90

Creator Teortaxes says frontier AI labs spend up to $15 billion each on highly inefficient training dataA browser-use dataset for SAP reportedly costs $500,000.

23h|113.5k491240

SH#1488|@HAMANDCHEESE

CR#1444|@CWOLFERESEARCH

RO#1220|@KRISHNANROHIT

FB#1153|@XEOPHON

CP#1092|@CHRISPAINTERYUP

SS#1073|@SYTELUS

KA#488|@YACINEMTB

T(#420|@TEORTAXESTEX

100

DeepSWE benchmark data shows GPT-5.5 outperforms Claude Opus 4.8 on software engineering tasks and token efficiencyClaude Opus 4.8 cost $12 per task.

13h|125.3k1.6k147

V(#1853|@REACH_VB

CH#1496|@KIMMONISMUS

LA#980|@SCALING01

AC#241|@_AIDAN_CLARK_

110

DeepSWE benchmark finds GPT-5.5 beats Claude Opus 4.8 on long-horizon coding with a 70% pass rate at half the costGPT-5.5 completed tasks in 21 minutes versus Claude's 43.

13h|125.3k1.6k147

V(#1853|@REACH_VB

CH#1496|@KIMMONISMUS

LA#980|@SCALING01

AC#241|@_AIDAN_CLARK_

120

Monzo co-founder Tom Blomfield argues replacing workers with AI causes operational chaos because models lack tacit domain knowledgeHe equates deploying AI to hiring capable but clueless workers.

11h|236.6k2.2k747

CE#1598|@INTUITMACHINE

⿻A#361|@IAMTRASK

GT#270|@GARRYTAN

130

Tom Blomfield, Y Combinator partner, argues replacing workforces with AI fails because structuring human domain knowledge is too difficultHe likens context-poor AI to hires lacking company context.

11h|236.6k2.2k747

CE#1598|@INTUITMACHINE

⿻A#361|@IAMTRASK

GT#270|@GARRYTAN

140

Luma AI Chief Scientist Jiaming Song, who led training for the Dream Machine video generator, has departedSong helped drive Luma's transition to multimodal foundation models.

5h|30.7k39726

NR#1562|@NATANIELRUIZG

ML#1449|@LIU_MINGYU

GO#1300|@GOWTHAMI_S

JG#674|@THOMA_GU

JS#509|@BAAADAS

MN#467|@MO_NOROUZI

AG#246|@ANIMESH_GARG

SX#158|@SAININGXIE

150

Bill Gurley and the All-In Podcast panel debate shifting AI job-displacement rhetoric and potential government threats to open-source modelsThe panel also analyzed Anthropic's advanced AI ambitions.

1d|1M2.9k1.6k

CE#1598|@INTUITMACHINE

CP#747|@CHAMATH

160

Corgi Insurance founder Nico Laqua details the extreme seven-day workweek culture powering the startup's $2.6 billion valuationTwo-thirds of early employees got company logo tattoos.

15h|560.2k1.7k565

B(#1990|@BRENDANFOODY

HS#1845|@HARRYSTEBBINGS

RO#1220|@KRISHNANROHIT

WM#1039|@WILLMANIDIS

B(#839|@BEFFJEZOS

170

Wharton statistics professor Weijie Su joins OpenAI to focus on model training while on academic leaveSu was simultaneously promoted to full professor at Wharton.

1d|28.1k19211

WS#1502|@WEIJIE444

AL#1383|@AARON_LOU

SL#1289|@SCOTT_LINDERMAN

DA#1271|@DAMEKDAVIS

JG#674|@THOMA_GU

YB#406|@YUBAI01

DP#197|@DIMITRISPAPAIL

JL#123|@JASONDEANLEE

180

Wharton statistics professor Weijie Su joins OpenAI on academic leave to train AI modelsThe move coincides with Su's promotion to full professor.

1d|28.1k19211

WS#1502|@WEIJIE444

AL#1383|@AARON_LOU

SL#1289|@SCOTT_LINDERMAN

DA#1271|@DAMEKDAVIS

JG#674|@THOMA_GU

YB#406|@YUBAI01

DP#197|@DIMITRISPAPAIL

JL#123|@JASONDEANLEE

190

Jiaming, who led Luma AI's pivot from 3D to video generation, has departed the company after three yearsHe steered the startup toward native multimodal foundation models.

5h|6.8k421

NR#1562|@NATANIELRUIZG

SS#1471|@_SAM_SINHA_

ML#1449|@LIU_MINGYU

AL#1383|@AARON_LOU

GO#1300|@GOWTHAMI_S

JG#674|@THOMA_GU

JS#509|@BAAADAS

MN#467|@MO_NOROUZI

AG#246|@ANIMESH_GARG

SX#158|@SAININGXIE

200

OpenAI's Thibault Sottiaux teases a GPT-5 roadmap where sequential decimal increments through GPT-5.5 deliver efficiency and speed gainsThe efficiency gains target costs in long-running agentic workflows

23h|663.5k5.1k404

CH#1496|@KIMMONISMUS

TI#768|@THSOTTIAUX

ER#306|@ERICMITCHELLAI

BB#133|@BOAZBARAKTCS

210

Claude Opus 4.8 scores 58% on DeepSWE coding benchmark, trailing GPT-5.5's 70% despite reducing average task costs to $12.58Chris Hayduk notes the model ran twice as slow.

12h|423.2k1.9k317

EL#475|@OMARSAR0

220

Claude Opus 4.8 scores 58% on the DeepSWE benchmark, trailing GPT-5.5 while cutting average task cost to $12.58It improved on Opus 4.7 by six percentage points

12h|423.2k1.9k317

EL#475|@OMARSAR0

230

Mullvad VPN's new London Underground ad campaign warns of potential UK government restrictions on VPN servicesPosters list countries banning VPNs, ending with 'United Kingdom?'.

22h|#1VIEWS2.3M#1LIKED43k#1BOOKMARKED2.1k

N🔎#1902|@NATHANPMYOUNG

240

Epoch AI finds open-weight AI models lag behind proprietary systems by an average of four monthsThe lag represents an eight-point deficit on Epoch's index.

1d|99.5k90879

XL#1674|@XLR8HARDER

GL#1480|@GLEECH

LA#980|@SCALING01

JC#251|@JULIEN_C

EM#175|@EMOLLICK

250

An overheard anecdote claims an incoming professor plans to spend her $1.5 million fund on ChatGPT tokens and undergraduatesThe strategy aims to outperform traditional graduate-led research workflows

23h|130.9k1.3k184

SG#1498|@SHARONGOLDMAN

TY#233|@TYPEDFEMALE

TL#224|@TALLINZEN

260

An overheard anecdote says a professor will spend a $1.5 million discretionary fund on ChatGPT tokens instead of graduate studentsUndergrads will feed open research problems into the model

23h|130.9k1.3k184

SG#1498|@SHARONGOLDMAN

TY#233|@TYPEDFEMALE

TL#224|@TALLINZEN

270

Anthropic's Opus 4.8 yields a 6% gain on DeepSWE benchmarks but still trails OpenAI's GPT 5.4The updated model reduces the average cost per task

9h|173.2k1.7k187

CH#1496|@KIMMONISMUS

KA#488|@YACINEMTB

280

Opus 4.8 improves DeepSWE benchmark performance by 6% over Opus 4.7 while lowering task costsOpenAI's GPT 5.4 continues to outperform the new model.

9h|173.2k1.7k187

CH#1496|@KIMMONISMUS

KA#488|@YACINEMTB

290

Google requests EPA approval to release 32 million Wolbachia-infected mosquitoes in Florida and California to suppress disease-carrying populationsPrior trials in Singapore reduced dengue cases by 70 percent.

11h|94.5k558103

VI#1195|@VIKHYATK

SK#505|@SEBKRIER

KA#488|@YACINEMTB

300

Michael Dell says Dell and Nvidia delivered the first Vera Rubin NVL72 system to CoreWeave, passing L11 diagnosticsThe delivery occurs while Blackwell optimizations are still transitioning.

4h|450.7k3.9k346

T(#420|@TEORTAXESTEX

RS#321|@SCOBLEIZER